Update TensorBoard training metrics

benchmark_cache/shard_scheduler.json

@@ -22,8 +22,8 @@
       "tokens_per_second": 598315.063954628,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 76749.99569587468,
-      "required_pages_per_second_upper_estimate": 0.005309742786765884
+      "required_token_positions_per_second": 76729.90254916946,
+      "required_pages_per_second_upper_estimate": 0.0053083526962542935
     },
     {
       "source": "rewrite6",
@@ -36,8 +36,8 @@
       "tokens_per_second": 8533.957213384612,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 9593.749461984335,
-      "required_pages_per_second_upper_estimate": 0.004016702559462623
+      "required_token_positions_per_second": 9591.237818646183,
+      "required_pages_per_second_upper_estimate": 0.004015650986846016
     },
     {
       "source": "openmath",
@@ -50,8 +50,8 @@
       "tokens_per_second": 380144.35761751566,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 19187.49892396867,
-      "required_pages_per_second_upper_estimate": 0.003996250017488459
+      "required_token_positions_per_second": 19182.475637292366,
+      "required_pages_per_second_upper_estimate": 0.003995203799346763
     },
     {
       "source": "ultra_qa",
@@ -64,8 +64,8 @@
       "tokens_per_second": 43742.16556775435,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 28781.248385953004,
-      "required_pages_per_second_upper_estimate": 0.0017253918943161775
+      "required_token_positions_per_second": 28773.71345593855,
+      "required_pages_per_second_upper_estimate": 0.0017249401867669828
     },
     {
       "source": "ultra_style",
@@ -78,8 +78,8 @@
       "tokens_per_second": 22018.86276523721,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 28781.248385953004,
-      "required_pages_per_second_upper_estimate": 0.0022376154666752033
+      "required_token_positions_per_second": 28773.71345593855,
+      "required_pages_per_second_upper_estimate": 0.002237029659009234
     },
     {
       "source": "cortex",
@@ -92,8 +92,8 @@
       "tokens_per_second": 910315.184299418,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 23,
-      "required_token_positions_per_second": 28781.248385953004,
-      "required_pages_per_second_upper_estimate": 0.3691749513981735
+      "required_token_positions_per_second": 28773.71345593855,
+      "required_pages_per_second_upper_estimate": 0.3690783014063256
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,24 +158,24 @@
     "prefetch_batches": 63,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 704566,
-    "producer_batch_seconds": 0.0962612154609701,
-    "producer_p95_seconds": 0.2866374250006629,
-    "gpu_update_seconds": 0.960625461037816,
-    "producer_batches_per_second": 10.388399888898745,
-    "gpu_batches_per_second": 1.040988439885453,
-    "producer_headroom": 9.979361432718553,
+    "producer_batch_seconds": 0.3823351382652902,
+    "producer_p95_seconds": 0.309508428999834,
+    "gpu_update_seconds": 0.9608770186141471,
+    "producer_batches_per_second": 2.615506397181134,
+    "gpu_batches_per_second": 1.0407159091412959,
+    "producer_headroom": 2.513179989089638,
     "sustainable": true,
-    "buffer_seconds": 60.51940404538241,
+    "buffer_seconds": 60.535252172691266,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 85,
+    "upload_pages_per_commit": 80,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 406594.19893153786,
+    "generation_bytes_per_second": 381070.3135667445,
     "upload_bytes_per_second": 5407499.924874017,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 87.24565006217577,
+    "estimated_full_shards_per_hour": 81.7688184285331,
     "selection_policy": "Smallest shard within 90% of measured local peak, enlarged to amortize 30 seconds of measured page production; bounded to 256 MiB."
   },
   "upload_measurements": [

metrics/void-byte-v1/1790868774355361490-de221e1d/train.jsonl

@@ -284,3 +284,4 @@
 {"step": 468600, "loss": 0.5792874097824097, "lr": 0.0003, "grad_norm": 0.14939919336839286, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 68589082129, "tokens_per_second": 151018.01464970456, "data_wait_s": 1.4115998055785894e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 468700, "loss": 0.5686940550804138, "lr": 0.0003, "grad_norm": 0.16503529398332553, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 68603781220, "tokens_per_second": 152822.21311803354, "data_wait_s": 1.3570999726653099e-05, "gpu_peak_gb": 89.37899398803711}
 {"step": 468800, "loss": 0.5900110602378845, "lr": 0.0003, "grad_norm": 0.1703935475430773, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 68618557387, "tokens_per_second": 153194.2599402319, "data_wait_s": 9.114002750720829e-06, "gpu_peak_gb": 89.37899398803711}
+{"step": 468900, "loss": 0.6242892742156982, "lr": 0.0003, "grad_norm": 0.16900648510523816, "recovery_retries": 0, "loss_scale": 1.0, "microbatch": 90, "input_bytes_and_specials": 68633333751, "tokens_per_second": 153725.33268365407, "data_wait_s": 1.4039000234333798e-05, "gpu_peak_gb": 89.37899398803711}

runs/void-byte-v1/events.out.tfevents.1790868774.f92f6aaf-743a-4c24-a54a-850bd75954cb-8js8b.612.0.1790868774355361490-de221e1d

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c2fc9f43b554da3d1c2e2013dd9f2901873deee9081256b1c9a6d625062f1b53
-size 609828
+oid sha256:c866672b97c2b5769b5937ab239a7ef666aaa509f030ec3b8ce32d4b299639c7
+size 611946